tarekziade's picture
tarekziade HF Staff
Link the tmic shrink page
db6238b verified
|
Raw History Blame Contribute Delete
2.49 kB
metadata
tags:
  - mic
  - test-artifact
base_model: peft-internal-testing/tiny-clip-text-2

tiny-clip-text-2-mic

Slim model weights, executable usage and framework expectations in one repository.

Contract

CLIP text encoder with projection

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: tiny-clip-text-2-mic

A 0.1M-parameter fixture for the CLIPTextModelWithProjection MIC, covering peft-internal-testing/tiny-clip-text-2. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

  • Fixture key 939edfa8be9f. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
  • Same architecture, smaller config: CLIPTextModelWithProjection, shrunk by the generic rule in sweep.py (num_hidden_layers 2).
  • Processor files copied byte for byte from peft-internal-testing/tiny-clip-text-2: merges.txt, special_tokens_map.json, tokenizer.json, tokenizer_config.json, vocab.json.
  • Random weights (seed 17). They passed the input-sensitivity gate: every perturbation of the contract's inputs moved the logits by at least 0.001 (weakest: attention_mask:all_ones, 1.44), so no training was needed.
  • Same tensor names as the real checkpoint. The real checkpoint has every parameter, and so does the fixture.
  • Saved in float32, the dtype peft-internal-testing/tiny-clip-text-2 loads in by default.
  • Finite in float32, bfloat16, and float16 (forward, loss, gradients).

Built by build_fixtures.py; build.json records the pinned sources, gate results, and file hashes. This repository includes the slim weights, usage and framework expectations; rebuild the other files with python example/build_fixtures.py --fixture tiny-clip-text-2-mic.